Methodological Evaluation and Time-Series Forecasting for Yield Improvement in Senegalese Transport Maintenance Depots

{ "background": "Transport maintenance depots in many developing economies face systemic inefficiencies, leading to suboptimal asset availability and high operational costs. In the Senegalese context, a lack of robust, data-driven methodologies for performance forecasting and yield analysis within these depots hinders targeted improvement initiatives. ", "purpose and objectives": "This study aims to methodologically evaluate current maintenance depot systems and to develop a predictive time-series model for forecasting yield, defined as the ratio of productive maintenance hours to total available hours. The objective is to provide a tool for evidence-based planning and resource allocation. ", "methodology": "A hybrid methodology was employed, integrating a structured evaluation of depot workflows with statistical modelling. Operational data from multiple depots were analysed. The core forecasting model is a seasonal autoregressive integrated moving average (SARIMA) process, formally specified as \ (B) \ (Bˢ) (1-B) ᵈ (1-Bˢ) D yt = \ (B) \ (Bˢ) \, where yₜ is the yield at time t. Model parameters were estimated using maximum likelihood. ", "findings": "The methodological evaluation identified critical bottlenecks in parts procurement and technician allocation. The SARIMA (1, 1, 1) (0, 1, 1) 7 model provided the best fit, forecasting a significant yield improvement of approximately 18% (95% CI: 14. 2% to 21. 8%) over a six-month period under optimised resource scenarios. Forecast uncertainty, measured by the mean absolute scaled error, was 0. 32. ", "conclusion": "The proposed time-series model offers a statistically sound and practically applicable tool for forecasting maintenance depot performance. It demonstrates that systematic data analysis can uncover substantial efficiency gains within existing operational constraints. ", "recommendations": "Depot managers should adopt formal time-series forecasting for capacity planning. Implementing a centralised data logging system is recommended to enhance model inputs. Further research should integrate real-time sensor data for predictive maintenance scheduling. ", "key